Originally Published MDDI November 2001

How Small is Small?: A Guide to the New Microfabrication Design and Process Techniques

Medical device manufacturers need to catch up on the rapidly evolving field of microfabrication to determine which processes and components are right for their product.

Bill Evans and Robert Mehalso

The term microfabrication is usually defined as the manufacturing of components and devices that are measured in hundreds of microns and that have tolerances of only a few microns. The term also applies to larger parts and assemblies with features that are measured in microns.

So just how small is a micron? Formally, a micron (µm) is one-millionth of a meter; 25.4 µm make up one-thousandth of an inch. The human hair makes its entry at 50 to 100 µm in diameter. It would not be unusual for the tip of a catheter to be less than a millimeter, which is 1000 µm.

But what is a micron in a more tangible way? Though it may sound implausible, many of the dimensions discussed in this article are actually bigger than many realize. People often think of the wavelength of light as being very small. Actually, it is barely "submicron," as the microchip manufacturers would say. Red light has a wavelength of 0.65 µm, or 650 nanometers (nm). Most people routinely handle low-cost products that have features injection molded into them; those feature dimensions are smaller than the wavelength of light.

The rainbow patterns on the common audio CD are created by data pits that are about a half a micron wide, but only an eighth of a micron deep (125 nm). These subwavelength features create a diffraction grating that splits light into its constituent colors as the disc is viewed.

But the manufacturing techniques used in popular consumer products like the CD are not just for the companies with giant R&D and tooling budgets. These techniques—precision machining, micromolding, electrical-discharge machining (EDM), laser processing, and a range of lithographic and electrodeposition methods—can all be used to help medical companies create very small or high-precision parts and assemblies to solve a variety of design and manufacturing problems.

Several technical and medical advancements are increasing the need for very small, close-tolerance parts and assemblies:

Engineers who are used to thinking in terms of conventional machining, where a tolerance of one-thousandth of an inch—a mil—is considered very tight, need only shift their thinking from mils to microns (about 25 times smaller) and understand that the milling machine or lathe is merely a different machine.

This article is intended for readers who are contemplating the design and manufacture of small parts or mechanisms but lack experience with microfabrication techniques. It focuses on techniques that design engineers in the medical device manufacturing industry can consider using today without too much experimenting and lab work. This article is not about micro- electromechanical systems (MEMS), which to many are synonymous with microparts or nanotechnology. MEMS are generally considered to be micromechanisms etched in silicon and other types of semiconductor materials, and it would require a separate article to describe their fabrication.

SELECTING MICROFABRICATION CANDIDATES

Tables I and II give an overview of the major processes and the materials to which they apply, and suggest potential applications. To determine if a part or assembly is a candidate for microfabrication, manufacturers first need to understand the big picture of the various processes and the geometries they can create.

The processes fit broadly into two categories. Mechanical approaches are the most familiar to those accustomed to creating larger parts. Light and chemical methods take advantage of processes originally developed for the semiconductor industry, as well as high-powered lasers. A common aspect of all vendors who offer these services is that they invest heavily in sophisticated measuring equipment to help control quality. This is known as metrology.

Table I. Overview of how microfabrication processes apply to various materials.

MECHANICAL PROCESSES

Micromachining and micromolding are similar in concept to their nonmicro brethren. A good rule of thumb says that if a design looks like it could be machined or molded if it were larger, then it can probably have the micro version of that process applied to it.

Micromachining. Highly specialized vendors usually perform micromachining with dedicated machining centers. These centers must be housed in controlled environments and are designed to work at significantly closer tolerances and smaller part sizes. Trying to work at tight tolerances by tweaking conventional machines is unlikely to achieve consistent results. Tools and jigs become very important because maintaining good surface finishes and actually holding such small parts become significant factors as size drops.

Micromolding. Micromolding requires both dedicated micro tooling skills and dedicated precision molding machines in a well-controlled environment. Successful micromolding requires two things: the equipment must intrinsically have the required resolution, and the vendor must exercise rigorous control over the materials, processing, and molding itself.

Molding and tooling for small parts are not just smaller versions of their regular molding counterparts, so manufacturers should seek vendors who have already demonstrated their skills on similar systems. Tooling is often created using electrical-discharge machining (see below) or diamond turning. It can be created with surface features below the wavelength of light by using lithographic and electrodeposition techniques (see below).

Variations on micromolding include microembossing, during which a lithographically produced master is typically pressed into polymers in a manner similar to compression molding. Elastomers such as silicone rubbers can also be molded in special equipment designed for processing thermoset plastics.

Electrical-Discharge Machining. EDM is commonly used in toolmaking for injection molding, but it can also be used to create individual parts. It is the erosion of metal by spark discharge—think of it as nibbling away at metal with tiny spark teeth that apply such little force to superhard materials that all the vibration and part-holding issues are greatly simplified.

There are two main types of EDM, wire EDM and electrode EDM. In wire EDM, a very thin wire electrode cuts a 2-D numerically controlled path in a metal blank in a manner similar to a tiny bandsaw. The blank sits in a bath of electrolyte and the wire has to enter and leave the piece just as a bandsaw blade would.

In electrode EDM, a male version of the desired machined cavity (the electrode) is first made and then sunk into the metal block. As a result, very sophisticated 3-D surfaces and shapes can be created. But because the electrode still has to enter and leave the metal block, re-entrant geometries are not usually possible. Electrodes are usually CNC machined in soft graphite or other soft metals for manufacturing ease, but for very-high-precision EDM the electrodes can be shaped with some of the precision light-based techniques outlined below.

In EDM, precision is achieved by two main factors. First, the type of machine used has to be designed for the desired tolerance and operated by someone able to exploit its accuracy; second, in electrode EDM, the resulting part can never be more accurate than the method used to create the electrode.

It is also possible to use slight variations on these themes to broaden the geometric possibilities. For instance, some vendors create rotating jigs to hold cylindrical workpieces in the EDM bath and then manipulate both the workpiece and the wire to cut features around the circumference. Some new wire EDM machines allow variable control of the wire orientation to create 2.5-D shapes.

Light and chemical processes either use light to directly remove material by high-energy laser vaporization (ablation) or to expose a photosensitive material (a resist) through a mask. The resist is applied to the substrate to a thickness of many mils. The exposed resist is then etched away, giving a precise high-aspect-ratio pattern. A conductive substrate with resist patterns can be put into an electrodeposition bath and metal parts can be "grown."

The accuracy of the processes is affected by several factors, most involving the wavelength of the exposing radiation or how the light is manipulated (masks, beam manipulation, etc.). Visible light can be used for exposure resolutions in the micron range, but UV light, electron beams, and even x-rays (LIGA process) can be used to get finer details and to produce thick high-aspect-ratio structures.

Collectively the photosensitive masking processes are often referred to as photolithography. In principle the processes are similar to the lithographic techniques used to print these very words. (Note that all of the processes described below can be performed using sheets or wafers, which are then sliced up to yield many individual parts.)

Lasers. Lasers are effective in micromachining all types of materials. It is important that the material being machined absorbs the laser energy. For instance, excimer (UV) lasers are used for metals, plastics, and ceramics; CO₂ lasers for metals and plastics; and femtosecond lasers for all materials. The laser's power, duration of pulse, and beam manipulation affect accuracy and speed of machining.

For example, optical-lens and shutter systems can be used to ablate axisymmetric 3-D surfaces such as lenses on flat plastic sheets. A variation of this process is used to ablate new lenses onto human eyes in the now popular refractive surgery procedures. The excimer laser ablates a lens shape onto the cornea by exposing it to several rapid pulses of decreasing diameter to sculpt the new optical prescription. In fact, the cornea material is very similar to a commonly used acrylic polymer, PMMA.

More-sophisticated beam shapers can be used to create more-complex 3-D shapes. Although the starting material may be a flat sheet, manufacturers are not limited to 2-D profiles. With the right equipment and in certain materials, the laser can be thought of as the cutter of a multiaxis milling machine. Not only can the 2-D profile be controlled very accurately and with no force applied, but the shape of the cutter can also be altered with lenses or mirrors. The depth of the cut can be controlled with focus or ablation time. This all adds up to the potential to create sophisticated micro 3-D parts in a wide variety of materials. Lasers are especially useful for small plastic parts that would be difficult to produce by other processes or would require costly and time-consuming tooling.

Lithographic Techniques in Silicon. Though originally developed for the semiconductor industry, lithographic techniques in silicon are now used routinely to create geometric parts with submicron accuracy. Some examples are miniature v-grooves, arrays of wells, and tiny gears. Because the substrate is silicon, the technique has the advantage of being able to combine small physical parts and features with electronic circuits to create integrated sensors and powered mechanisms. When combined with electrical function, such parts become MEMS.

Although MEMS fabrication is often regarded as an exotic technology, manufacturers should not dismiss silicon etching. It is useful for the cost-effective manufacture of even small quantities of superaccurate parts.

The process starts with a standard-diameter thin wafer of silicon sliced from a giant cylindrical crystal. A photoresist is coated onto the surface and the desired pattern is exposed onto it through a mask, then developed. The wafer is next immersed in a chemical bath (typically sodium or potassium hydroxide), and the patterned silicon is etched. The etching takes place along the planes of the crystal.

Silicon crystals are available in various crystallographic orientations that result in different etch geometries. Etching along crystallographic planes is useful in creating superaccurate 2.5-D structures in silicon. This is sometimes referred to as crystallographic etching. Not only is the 2-D profile accurate to a tolerance of about a micron, but the depth of a v-groove can also be accurately controlled.

Usually the desired parts are made in multiple quantities on the wafer—from hundreds to thousands—and then either scribed and snapped or diced using a rotary saw. This is known as wafer-scale batch processing and is an important reason small parts made this way can be so cost-effective. True, the process is expensive per batch, but when spread over hundreds of parts, cost per part becomes very low. Also, tolerances are good on each individual part and across the wafer diameter. Wafers consisting of thousands of parts can be stacked and bonded together to create more-complex geometries.

Lithographic Etching of Other Materials. It is possible to etch a wide variety of materials in a manner similar to silicon etching (see Table I for specific materials). It is also possible to use special resists or patterns that produce 3-D surfaces rather than just cutting a 2-D pattern straight down. For instance, on glass it is possible to etch microlenses this way with a kind of graduated tint-pattern mask exposed onto the surface. This causes the material to etch at different rates at different points on the diameter, creating a curved surface.

Electrodeposition. Electrodeposition is the manufacturing of parts by electrodepositing metals into a mold. Typically, for 2.5-D parts, precision features are made on the mold using lithographic techniques. For instance, hundreds to thousands of intricate, thin-walled metal parts can be made at once using a master mold. The surface of the conductive mold is passivated such that the precision electrodeposited parts can be released.

Complex 3-D parts are generally electrodeposited on a precision-machined substrate. The substrate or mold material is selected so that it will preferentially etch relative to the electrodeposited part. For example, an aluminum mold can be quickly etched away, leaving electrodeposited nickel parts.

SURFACE BUILD-UP TECHNIQUES

With the exception of electrodeposition, most of the above light-based processes are subtractive. It is also possible to add low-aspect-ratio features to the surfaces of microparts using the following techniques. Materials can be built up over the whole surface or in selectively masked areas. For instance, electrical wiring can be added to the surface of a ceramic part by depositing a thin conductive layer onto the surface and then electroplating a thicker layer of gold onto the tracks, making them capable of carrying meaningful current.

Physical and Chemical Vapor Deposition. Physical vapor deposition (PVD) and chemical vapor deposition (CVD) are processes for coating materials onto surfaces. They are typically done in a vacuum chamber. With PVD, many materials can be vaporized with heat or can be "sputtered" onto surfaces with great control over the resultant depth. CVD is similar to PVD but uses a chemical reaction in the vapor or at the surface to deliver and bond new materials. Typically, CVD is useful in the production of many metals and particularly ceramic coatings. It is also possible to use lasers (laser CVD) to selectively bond materials in specific locations where the heat has triggered the necessary reaction.

MANAGING FROM CONCEPT TO PRODUCTION

Project Management. Most medical manufacturers are well aware that they must consider all aspects of design, manufacturing, and regulatory requirements at the early concept stages to avoid downstream problems. Microfabricated design solutions require that such efforts be increased tenfold.

Several of the processes detailed in this article are relatively immature and chances are that many medical designers are tackling small-part design problems for the first time. If the development team is not well informed, it may miss major opportunities to solve problems in an innovative, cost-effective, and time- efficient way. Companies may seek expert advice for their designers by hiring team members from industries with past exposure to micromanufacturing technologies, bringing in experienced consultants to jump-start the effort, and asking potential vendors for their input on possible solutions early in the design cycle.

Design Management. In the creation of microparts, which comes first, the design or the manufacturing method? The intelligent response is neither. Manufacturers should not design the part and then go looking for a process, nor should they pick a process and design within its restrictions.

Instead, the project should start with an open-minded and informed brainstorming process where all factors are thrown in. Manufacturers should lay out the design freedoms, tolerances, and functional requirements in a way all on the team can understand, then challenge them to be creative with solutions. The partially formed design possibilities can be brought to vendors for critical feedback.

Manufacturers should consider at least three possible processes, and hypothetically walk the potential solutions through all the stages necessary to get to market, soliciting feedback from all those who will need to be involved. This will generate valuable realistic comments on design, manufacturing, scheduling, and cost issues before the team is locked into the consequences of a hastily picked design solution.

Typically, brainstorming and early feasibility studies take about six to eight weeks, and are a great way of rounding up all the project issues. For very complicated designs or parts out on the leading edge of possibility, managers may also want to consider running several alternatives at least part of the way through the design and manufacturing cycle before settling on the final solution.

Design Tips. There is true power in making many parts at once on a sheet or wafer scale. The actual process steps may be expensive, but each batch yields thousands of parts. Sheet- or wafer-scale approaches also work well if building up a series of planar layers can create the design. Alignment between the layers can be better than ±5 µm. Each time a new layer is attached, conceivably thousands of features may have been added to the wafer. Re-entrant shapes are also possible with this approach; engineers must be cautious, however, to match thermal-expansion coefficients between layers of different materials. Caution is also warranted when using modeling and simulation tools such as finite element analysis and mold-flow analysis—they may not be accurate at such small dimensions.

Manufacturers must inspect and review prototypes and early production parts carefully. Scanning electron microscopy is vastly superior to optical alternatives, and is not overly expensive at service bureaus. Literally having the vision to check the progress is invaluable.

Prototyping. Prototyping is a significant problem, as there are rarely good surrogates for the final intended process. Most of the time designers will have to bite the bullet and make smaller batches using the intended technique and accept that it will be expensive and time-consuming.

To help make each prototyping stage more effective, manufacturers must not rush in. Instead, they should take the time for vendor feedback and consider making larger-scale models of the design with cheaper prototyping technologies (SLA, SLS, etc.). If the design is planar with many layers, then a large-scale model can be made quickly using low-cost conventional laser cutting of inexpensive plastic sheeting.

Prototyping in light- and chemical-based processes may be time-consuming, but it has the advantage that parts are produced using the tooling and equipment intended for final production. The process may need scaling up, but once it is proven, moving to production volumes can usually be done quickly.

Prototypes of parts intended for molding pose special problems, as plastics are much harder to machine at very small dimensions. Laser machining the prototypes in the intended materials, or making initial samples in metal, can help prove the geometry at the correct scale.

Vendor Selection. To some manufacturers, a particular geometry might look impossible at first, but somewhere in the world there are probably one or two vendors doing something similar. Manufacturers should look around at other industries or market segments to see who is making something similar to the intended design.

For example, a company was recently attempting to mold a high-aspect-ratio (10:1) capillary tube with an inner diameter of 500 µm for a biotech fluid-handling instrument piece. This appeared to break all the rules of molding, but upon investigation, similar aspect ratios were being achieved in laboratory pipetting systems. A search of these molders revealed two vendors in the United States able to mold the part.

Caution should be taken when using the Web for vendor searches, however. The microfabrication industry is not yet mature. Web sites are often small, obscure, and poorly indexed by search engines, and terminology is not standardized. These factors conspire to make most Internet searches incomplete. Companies are better off using their networks, consultants, and trade publications and associations to probe for vendors.

Manufacturers can lower the risks of vendor choice by taking unconventional approaches. For instance, when choosing between two vendors, how can a project manager objectively determine which is most likely to succeed if neither vendor has been successful before? Both vendors may be equally committed to working towards a solution. Before deciding, collect existing parts created by these vendors that are close to the desired new part size, geometry, or special design issues. Take these parts and subject them to independent testing and inspection. Comparing impartial scanning electron microscope images or structural test reports can help determine who is most likely to succeed.

Manufacturing. When considering the final manufactured cost of a microsystem, be aware that the core microcomponents themselves comprise a relatively small proportion. These may only account for 10 to 20% of the cost, with the balance being 15–25% for measuring and testing during production and from 55 to 75% for component handling, packaging, interconnection, and assembly to other components. Therefore, it is crucial that the design team is working together with the manufacturing team; successful, cost-effective systems will come from optimizing this relationship.

The act of designing any product also causes the design of its manufacturing system—this is especially true for microsystems. A precision active alignment of a few microns during the assembly of a number of parts might force the use of very costly robots or sophisticated inflexible jigs that have to be remade every time required dimensions change. If a similarly functioning design could be created using intrinsically close-tolerance, self-aligning parts (crystallographically etched silicon done at wafer scale by the thousands and then routinely diced, for example), much of the burdensome manufacturing infrastructure and long lead times could be eliminated.

Assembly. Handling small parts can be a difficult issue and should be considered from the beginning. It is likely that features will have to be designed onto the part to facilitate handling at various times, from holding in jigs on machine tools to actually manipulating a part into place by the intended end-user. Forces that are normally small—such as static, thermal, and fluid surface tension—can become big factors.

The fasteners that engineers normally use are avoided in micro-systems because they limit miniaturization and create stress problems. Attachment techniques such as soldering, diffusion and anodic bonding, the use of adhesives (organic and nonorganic), and sonic and laser welding are preferable. It is normal to mix and match such methods in an assembly. Manufacturers should avoid active alignments if possible or do them at wafer scale. Designers must understand and apply the first principles of alignment and issues of overrestraint to avoid building stress and misalignment into the assembly.

If possible, designers should try to leverage equipment that exists in another industry to assemble the product. The semiconductor industry routinely uses robots to put together parts at 5- to 10-µm alignment; even better alignment is possible with the newer machine vision–guided systems.

There is a technique in chip assembly that can align parts to ±2 µm using the surface tension of solder. It may be possible to apply this approach to a micro medical system problem. Gold pads are laid down lithographically onto the target substrate and then coated with solder paste. The chip that is to be aligned and connected with these pads has similarly accurate gold pads also lithographically applied. The chip is placed above the solder pads at ±15-µm accuracy using conventional robots. When the solder is turned to liquid during attachment, the meniscus surface-tension effect balls the solder exactly onto the center of the superaccurate gold pads. The chip then "floats" into alignment that nearly equals the lithographically etched accuracy of the pads. Engineers may not need to be making an electrical connection to exploit this effect in other applications.

CONCLUSION

It will likely cost more and take longer than expected in the front-end design and prototyping to implement a micropart or microassembly. On the bright side, however, the microprocesses can often be directly transferred to production. Therefore, the overall product implementation schedule may be faster and less costly.

Adopting a "get-to-market-right-the-first-time" approach to project management by paradoxically checking often for potential problems is crucial. Mistakes will be made on the ultimate path to success. The key is to prototype, model, create small batches, test, validate, and inspect the design as it moves forward. Iterate, iterate, iterate; seek outside experienced advice; accept mistakes, learn, and move on before committing to final production. That way, the cost-effective product will make it to market right the first time, on schedule.

Bill Evans is founder and principle of Bridge Design Inc. (San Francisco); Robert Mehalso is senior vice president and chief technology officer at Ardesta LLC (Ann Arbor, MI).

Photo Courtesy of Robert Mehalso

Copyright ©2001 Medical Device & Diagnostic Industry